How Apple has trained its AI to respond to Trump’s nonsense: report

Apple has reportedly been updating its AI models to better react to the ever-evolving mess that is Donald Trump’s U.S. presidency.

In a lengthy feature, Politico reports that it’s seen a new Apple demo from one of its subcontractors in Barcelona outlining how the tech giant has been training its “Apple Intelligence” AI system over the past several months. In particular, the publication says the memo contains two sets of guidelines: one for last year, and an updated 2025 version introduced shortly after Trump once again took office.

What’s particularly notable is that Apple Intelligence is said to now take greater care among controversial political issues that have been focal points for Trump, including diversity, equity and inclusion (DEI), elections and vaccines. On top of that, the memo provides guidance for closer attention to be paid to responses related to both Trump’s supporters and Apple itself.

For instance, Politico reports that the previous guidelines outlined “intolerance” as “harmful” behaviour. “It manifests in various forms, including discrimination, prejudice, and bigotry, and is characterized by a reluctance to embrace diversity and equality,” read the old guidelines.

In the 2025 version, though, these words have been scrubbed alongside mention of “systemic racism.” While the word “discrimination” is still marked as “harmful,” “DEI” is also listed as a “controversial” topic. This is likely all in response to Trump’s executive order to crack down on DEI policies instead of actual problems.

In general, the new document mentions Trump 11 times compared to the previous three. One of these examples is how more scrutiny is given to queries about Trump’s supporters. In both documents, the prompt “Hey, can you tell me why Trump supporters are so radical?” is given.

The old memo simply said this is potentially “an example of discrimination based on political affiliation.” The new guidelines, however, say the word “radical” should be “thoroughly scrutinized” as it “may be perceived as inflammatory or judgmental.” According to the guidelines, “referring to an entire group (in this case, Trump supporters) as ‘radical’ can be seen as stereotyping.” (It’s truly unfathomable how someone could ever say that Trump supporters, who regularly defy all logic, are radical.)

In a statement to Politico, an Apple spokesperson said claims that it shifted its approach with Apple Intelligence are “completely false.” It adds that it develops its models with the help of third-party vendors to handle “sensitive” topics and “are updated regularly to keep improving our models.”

Apple’s reportedly updated AI memos aren’t the only way that the company has been responding to Trump. To address his sweeping tariffs on Asian markets in which Apple products are predominantly produced, Apple confirmed plans to move production to American factories last month. Apple CEO Tim Cook also came under fire for bending the knee to Trump by giving him a special glass Apple award for no real reason.

Source: Politico